Linux I2C development

* Re: [PATCH 06/10] i2c: i2c-mux-ptxpmb-cpld: Add device tree bindings
From: Peter Rosin @ 2016-10-11  7:36 UTC (permalink / raw)
  To: Rob Herring, Pantelis Antoniou
  Cc: Mark Rutland, Alexandre Courbot, Georgi Vlaev, linux-watchdog,
	Wolfram Sang, devicetree, Frank Rowand, Linus Walleij,
	Guenter Roeck, linux-kernel, JawaharBalaji Thirumalaisamy,
	linux-gpio, Wim Van Sebroeck, linux-mtd, Debjit Ghosh, Rajat Jain,
	Brian Norris, Lee Jones, David Woodhouse, Guenter Roeck,
	linux-i2c
In-Reply-To: <20161010174517.GA18097@rob-hp-laptop>

On 2016-10-10 19:45, Rob Herring wrote:
> On Fri, Oct 07, 2016 at 06:17:27PM +0300, Pantelis Antoniou wrote:
>> From: Georgi Vlaev <gvlaev@juniper.net>
>>
>> Add binding document for the i2c driver of PTXPMB CPLD.
>>
>> Signed-off-by: Georgi Vlaev <gvlaev@juniper.net>
>> [Ported from Juniper kernel]
>> Signed-off-by: Pantelis Antoniou <pantelis.antoniou@konsulko.com>
>> ---
>>  .../bindings/i2c/jnx,i2c-mux-ptxpmb-cpld.txt       | 50 ++++++++++++++++++++++
>>  1 file changed, 50 insertions(+)
>>  create mode 100644 Documentation/devicetree/bindings/i2c/jnx,i2c-mux-ptxpmb-cpld.txt
>>
>> diff --git a/Documentation/devicetree/bindings/i2c/jnx,i2c-mux-ptxpmb-cpld.txt b/Documentation/devicetree/bindings/i2c/jnx,i2c-mux-ptxpmb-cpld.txt
>> new file mode 100644
>> index 0000000..3b201f7
>> --- /dev/null
>> +++ b/Documentation/devicetree/bindings/i2c/jnx,i2c-mux-ptxpmb-cpld.txt
>> @@ -0,0 +1,50 @@
>> +* Juniper PTXPMB CPLD I2C Mux
>> +
>> +I2C mux on the PTXPMB CPLD on Juniper series of routers.
>> +
>> +Required properties:
>> +
>> +  - compatible: Must contain one of the following.
>> +    "jnx,i2c-mux-ptxpmb-cpld", "jnx,i2c-mux-ngpmb-bcpld"
>> +  - num-enable: Number of muxes to enable.
> 
> No. Use status property.

Reading the code, I understand this mux to be like a combination of
i2c switches (like pca9548) where the slave i2c busses can be
enabled individually using an "enable" bitmask in one register, and
"regular" i2c muxes (like pca9547) where you can only select one
slave i2c bus at a time using an enumeration.

num-enable would be the number of switches and num-channels would be
the number of muxes per switch channel. Or the other way around? This
is a real problem. The bindings are too tightly coupled to the way
Linux currently handles i2c switches (i.e. not very well, you can only
select one slave bus on an i2c switch, thus reducing the more
flexible switch hardware to a regular mux). The bindings should not
describe what the driver does, they should describe the hardware.

Basically, you need to describe the i2c topology of the hardware in
the bindings, because I'm just guessing all this, and I shouldn't
have to.

Cheers,
Peter

>> +
>> +  The following required properties are defined externally:
>> +
>> +  - Standard I2C mux properties. See i2c-mux.txt in this directory.
>> +  - I2C child bus nodes. See i2c-mux.txt in this directory.
>> +
>> +Optional Properties:
>> +
>> +  - num-channels: Number of channels. If not present the default is 8.
> 
> What's a channel?
> 
>> +  - base-bus-num: Base bus number. If not present it is 0.
> 
> No. Linuxism and why do you care?
> 
>> +  - use-force: Use the force method of the controller.
>> +
>> +
>> +Example:
>> +
>> +cpld-i2c-mux {
> 
> mux {
> 
>> +	compatible = "jnx,i2c-mux-ptxpmb-cpld";
>> +	#address-cells = <1>;
>> +	#size-cells = <0>;
>> +
>> +	i2c-parent = <&i2c1>;
>> +
>> +	num-enable = <1>;
>> +
>> +	i2c@0 {
>> +		reg = <0>;
>> +		#address-cells = <1>;
>> +		#size-cells = <0>;
>> +
>> +		/* PMB devices are accessed through FPC */
>> +
>> +		temp-sensor@1a { /* FPC */
>> +			compatible = "maxim,max6697";
>> +			reg = <0x1a>;
>> +			smbus-timeout-disable;
>> +			resistance-cancellation;
>> +			alert-mask = <0xff>;
>> +			over-temperature-mask = <0xff>;
>> +		};
>> +	};
>> +};
>> -- 
>> 1.9.1
>>

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

^ permalink raw reply